Machine learning potentials

Module that contains the class for evaluating the energy of molecules via machine-learned potentials through the ase package

class topsearch.potentials.ml_potentials.MachineLearningPotential(atom_labels: list, calculator_type: str = 'torchani', model: str = 'default', device: str = 'cpu')

Description

Evaluate the energy of a molecular system using machine learning potentials called through the ase package. Options are ANI2x and MACE.

atom_labels

List of atomic species for each atom in the molecule.

Type:

list

calculator_type

Type of calculator to use. Options are ‘mace’ or ‘torchani’

Type:

str

function(position: NDArray[Any, Any]) → float

Compute the electronic potential energy

function_gradient(position: NDArray[Any, Any]) → tuple

Compute the electronic potential energy and its forces

gradient(position: NDArray[Any, Any]) → NDArray[Any, Any]

Compute the analytical gradient from the ASE calculator